Table 2 Performance metrics from 10-fold cross validation for random forest classification of high and low function women on each of 11 feature sets
From: Accelerometer-based predictive models of fall risk in older women: a pilot study
Set | Accuracy | Precision | Sensitivity | F1-Score | AUC | Feature groups | Top-five features |
|---|---|---|---|---|---|---|---|
1 | 69.0% | 75.0% | 0.873 | 0.807 | 0.545 | Gait | STD_STEP_TIME, STD_STRIDE_TIME, MEAN_STRIDE_TIME, MEAN_STEP_TIME, CADENCE |
2 | 71.9% | 79.0% | 0.845 | 0.817 | 0.665 | X-axis | X_MEAN, X_COV, X_SMA, X_PFREQ, X_ENERGY |
3 | 72.7% | 79.1% | 0.858 | 0.823 | 0.661 | X-axis, gait | X_MEAN, X_COV, X_SMA, X_RMS, X_ENERGY |
4 | 75.9% | 82.6% | 0.855 | 0.840 | 0.730 | Y-axis | Y_PFREQ, Y_MCR, Y_COV, Y_STD, Y_MAD |
5 | 76.2% | 82.5% | 0.862 | 0.843 | 0.727 | Y-axis, gait | Y_PFREQ, Y_MCR, Y_RMS, Y_STD, Y_COV |
6 | 70.9% | 83.3% | 0.760 | 0.795 | 0.759 | Z-axis | Z_COV, Z_MEAN, Z_MAD, Z_STD, Z_PFREQ |
7 | 73.1% | 84.1% | 0.785 | 0.812 | 0.771 | Z-axis, gait | Z_COV, Z_MEAN, Z_MAD, Z_STD, Z_SMA |
8 | 70.9% | 79.3% | 0.822 | 0.807 | 0.616 | Vector magnitude | MAG_PFREQ, MAG_MAD, MAG_MCR, MAG_P2P, MAG_SMA |
9 | 71.4% | 78.5% | 0.846 | 0.814 | 0.616 | Vector magnitude, gait | MAG_PFREQ, MAG_MCR, MAG_MAD, MAG_MEAN, MAG_SMA |
10 | 79.3% | 84.6% | 0.881 | 0.863 | 0.834 | XYZ, cross-correlations | Z_COV, XZ_CORR, Z_MEAN, X_MEAN, Z_MAD |
11 | 78.9% | 84.4% | 0.877 | 0.860 | 0.846 | XYZ, cross-correlations, gait | Z_COV, XZ_CORR, Z_MEAN, X_MEAN, Z_MAD |
AVG | 73.7% | 81.1% | 0.842 | 0.826 | 0.706 |
